@xenoGitHub ↗

Jimmy T

Solo founder building most on Composer 2.5 and Grok 4.6. Using wave to manage projects, and Cursor CLI Agent harness with "agency agents" persona. Everything is routed through my delegator agent who splits out work to specialist subagents as needed. Automated e2e and unit tests.

Viewing version 2

Back to latest →

Rewrite public summary to match the actual Cursor harness, not a generic pairing line.

You’re reading the setup as it was published on 17 September 2026.

Cursor-first, same-chat implementation

What I use AI for

Product work in a coding agent: first implementation, small follow-ups, and a second look before I accept a change.

Agents & harnesses

  • Cursor Agent (current): Daily coding harness. I stay in the same chat unless I name a persona.
  • Composer 2.5, Grok 4.6 (current): The model I leave selected.

Apps

  • Wave: Wave is a multi terminal tool that lets me save terminal layouts as projects.
  • Cursor CLI: Generous subsidized tier for in house models maximizes my utility.

Skills

  • Project rules and named personas (current): Accessibility, copy casing, and domain rules travel with the work. A large specialist roster exists; I only invoke it by name.
  • Agency Agents: These guys work in specialty roles to accomplish targeted tasks.

MCP connections

  • Context7: Keeps my docs on track

How I work

  • I describe a small visible outcome in Cursor Agent and let that chat implement it.
  • I review the change; work stays uncommitted until I ask to commit or push.